Understanding the behaviour of rocks subjected to high temperatures and pressures is essential for a wide range of subsurface applications. This thesis aims to develop an image analysis methodology to measure the porosities and permeabilities of rock samples. These samples were created as part of the doctoral thesis of K.H.A.A. Wolf by compacting rock rubble in a copper tube and subjecting it to various high temperatures and pressures. Despite earlier image analysis on these samples, there was an opportunity for improvement using higher-resolution imaging and more advanced processing techniques. UsingaDSLRcameraand2typesoflight (normal and UV), 33 samples were photographed and then analysed using Adobe Photoshop and ImageJ. Additionally, three samples were selected to be scanned using a µCT scanner. The image processing included the conversion to greyscale, binarisation and the use of the morphological ”open” operation. Porosity was directly measured, while permeability was estimated using a modified Kozeny-Carman equation. The quality of the impregnation of the sample with UV dye differed, yet for 28 out of the 33 samples, imaging with UV light worked better than with normal light. Using temperature data from when the samples were formed, a clear trend emerged showing that the higher the temperature, the lower the porosity and permeability become. This trend was also visible in the work by Wolf (2006) and the values were generally in agreement, except for the samples with extremely low porosities and permeabilities. While CT and image analysis gave similar porosity values, permeability results differed, highlighting the need for simulation-based permeability estimates from CT data. Further research into impregnation techniques is also recommended to enhance future image analysis workflows.

Analysis of petrographic thin sections is crucial to determine rock properties such as texture, crystal size distribution, porosity and structure of rocks at a microscopic level. However, an accurate determination of minerals and textures in petrographic thin sections is a time-consuming process. Despite efficiency improvements through intensity-based segmentation, thin section petrography remains slow and manual, leaving a significant amount of sections unexamined due to time constraints. This project aims to determine whether object-based segmentation can be used to segment similar mineral phases in petrographic thin sections using the AI-driven ZEISS arivis Cloud (formerly APEER).

Samples used in this project were collected by the authors of Schmiedel et al. (2021) across a traverse of the high viscosity Sosa Dyke (37°2’S, 68°52’W) in Argentina. Ten thin sections of trachyte/trachydacite composition were scanned under bright light (BL), circularly polarized light (CPL), cross-polarized light (XPL) and plane-polarized light (PPL) using the ZEISS Axioscan 7 and provided to be used in this project.

The effects of annotation quantity, image resolution and the combination of CPL and BL images in a single model were investigated through the creation of multiple AI segmentation models using arivis Cloud. Additionally, the time spent on segmentation was compared to intensity-based segmentation, point counting, and manual segmentation. The classes used to segment images were amphiboles, plagioclase, and opaque minerals.

Results from this project show that ZEISS arivis Cloud can be used to segment thin sections using object-based AI models. Object-based segmentation with ZEISS arivis Cloud is shown to be more time-efficient than other manual and intensity-based segmentation methods. Larger datasets of similar thin sections result in greater time savings, and the platform also generates valuable data which can be used for geologic interpretation. BL images produced more accurate models than CPL images, and increased annotation of images improves model accuracy. Models trained on higher resolution images more accurately differentiated between plagioclase and amphiboles. Conversely, higher resolution models were less consistent in identifying large plagioclase crystals. Annotation bias and arivis Cloud performing optimally when objects are no larger than 320*320 could be reasons for this. Although contradictory to the recommendations of the arivis Cloud documentation, combining BL and CPL images into a single model may improve segmentation accuracy when applied to BL images.

If fully optimized, ZEISS arivis Cloud could become an important tool for analyzing thin sections that were previously deemed too time consuming, potentially unlocking new geologic insights

With demand for lithium predicted to continue increasing in the coming years, lithium exploration and the development of recovery techniques to extract it will become increasingly important. Lithium is particularly important for the energy transition due to its use in lithium-ion batteries. A major source of lithium are Lithium-Caesium-Tantalum (LCT) pegmatites. Lithium is contained within spodumene crystals and these are therefore the focus of most extraction techniques. The main objective of this study was to create an overview of the lithium distribution in lithium pegmatites based on textural and mineralogical information, combined with elemental analysis using Laser Induced Breakdown Spectroscopy (LIBS). The purpose was to improve the understanding of spodumene characteristics in lithium pegmatites which could affect lithium exploration and recovery.
Six samples from two different lithium pegmatite fields were analysed. Four samples were from the Bergby pegmatite field in Sweden and two samples were from the Bougouni pegmatite field in Mali. Petrographic descriptions were made to understand the mineralogy and textures of spodumene and LIBS analysis was used to measure lithium concentrations across profiles on a micrometre scale. Notable features in the Bergby (Sweden) samples included recrystallised, fine grained spodumene in uneven rims around the primary spodumene crystals as well as in the finer grained groundmass. The average lithium content measured for spodumene from the block samples was between 1.4 to 2.1 wt % Li, which is lower than the commonly measured range for spodumene of 2.8 to 3.5 wt % Li (Aylmore et al., 2018). LIBS analysis on the thin sections was attempted and showed higher average lithium content in the micrometre-scale fractures in the spodumene than in the surrounding spodumene crystal. The results of this thesis show that the majority of the lithium remains contained within primary spodumene crystals in lithium pegmatites but lower concentrations are also present in certain locations in the surrounding rock. Additionally, it indicates that the Keyence EA-300 may not be suitable for Li detection in spodumene crystals due to the significant differences in lithium concentration between thin section and block sample measurements. In future it would be critical to consider factors such as matrix effects, sample preparation and calibration of the Keyence EA-300 for true quantitative analysis of lithium content in spodumene.

The DAP well is a project by the TU Delft to be able to provide heat to a part of Delft. However, presently the geothermal well is not economically feasible. The well could be a potential source of lithium, which is mostly used in batteries. The demand for lithium is estimated to rise, thus this could potentially be a promising investment. In this research, the feasibility of extracting lithium from the clay and mica faces of the Delft Sandstone is estimated. The Delft Sandstone was formed in the Valanginian era and consists of sandstone and siltstone with clay interfaces. Potential minerals that could occur in this layer that can contain lithium are smectites, kaolinites, micas and illites, fibrous clays or chlorites. These minerals either contain lithium in their naturally occurring mineral structure or can be formed when in a mineral an element is substituted by lithium. The minerals that will be analysed in this research are hectorite, montmorillonite, bentonite, kaolinite, lepidolite, zinnwaldite, illite, sepiolite, palygorskite and cookeite.
The samples will be analysed and interpreted using Gamma-Ray Spectroscopy and Infrared spectroscopy. Gamma Ray Spectroscopy detects the naturally occurring gamma rays emitted from a rock sample. These are quantified in a general signal as well as concentrations of Potassium, Thorium and Uranium. High clay content will relate to a high K value thus a high GR (Gamma Ray) signal. In IR (Infrared) Spectroscopy a beam of infrared light is passed through a sample, which leads to the absorption of specific wavelengths by the molecular bonds present. These spectra are then compared to library spectra and minerals can be identified. In this thesis, the GR signal could not be related to the well logs as the sample quantity was too low. From the IR Spectra kaolinite and illite were defined in this layer. These clays can contain lithium when elements are substituted in their molecular structure.

The growing importance of rare earth elements (REEs) in sustainable technologies necessitates anefficient assessment of potential resource targets within the European Union. Traditional analyticaltechniques for REE determination have drawbacks like destructive and time-consuming samplepreparation, but hyperspectral imaging (HSI) and laser-induced fluorescence (LiF) offer promisingalternatives. This project aims to use a combined HSI and LiF method to qualitatively characterizeREE mineralization and alteration in drill core samples from the Storkwitz deposit, Germany. The goal is to develop a transferrable mapping approach for REEs, while enhancing our understanding of the Storkwitz deposit. It aims to check the robustness of HSI conducted across the extendedwavelength range as a tool in effectively characterizing the lithologies associated with REEmineralization in the Storkwitz breccia. It also tests if the combined HSI-LiF can provide newinsights into the presence of REEs and its associated minerals in the Storkwitz breccia.The proposed workflow involves acquiring hyperspectral data of the Storkwitz drill core and blocksections in three wavelength ranges (VNIR-SWIR, MWIR, and LWIR). Automatic and manualendmember extraction is performed on the smaller subsection of data to create spectral libraries,which are then used for spectral unmixing and mapping of the entire hyperspectral dataset toidentify lithologies. LiF is used to identify and map rare earth elements (REEs) in selected REErichzones identified from the hyperspectral data.The results indicated that the Storkwitz Breccia is primarily composed of ankerite in the matrix,along with minor amounts of white mica, clay, iron oxides, and REE-fluorcarbonates. The brecciacontains different clasts, including granitoid clasts rich in orthoclase and quartz-albite-biotite, aswell as carbonatitic clasts dominated by ankerite, dolomite, and ankerite-calcite. The breccia alsounderwent four alteration phases, including fenitization, biotite alteration, white mica-clayalteration, and ferric alteration. Laser-induced spectroscopy confirmed the presence of REEs, withapatite and REE-fluorcarbonates, particularly bastnäsite, being the main REE-bearing minerals.The specific REEs identified include Nd3+, Sm3+, Pr3+, and possibly Eu3+. The comprehensiveworkflow combining hyperspectral imaging and laser-induced fluorescence spectroscopy provedto be a successful approach for characterizing lithologies and mapping rare earth elementmineralization in the Storkwitz breccia. The study opens up new possibilities for efficient REEexploration in similar geological settings, providing valuable information for geological loggingand interpretation.

New mining and exploration projects have revived across Europe with an increasing demand for critical raw materials driven by the energy transition and unstable political conditions. Hyperspectral imaging proves to be an excellent tool for quick and efficient non-invasive mineral exploration. By mounting hyperspectral- and LiDAR sensors on drones and other robots, inaccessible and dangerous areas can safely be mapped for geological information. With increasingly deeper and more hazardous underground mines, a robot-aided
hyperspectral mineral exploration method in underground environments is needed. This study configures a Cubert X20P hyperspectral camera (VNIR) and VLP16 LiDAR on the versatile Boston Dynamics Spot robot, creating a multi-sensor robotic platform for data acquisition in underground mining environments. A data workflow is proposed and applied to the granite greisen rocks of the Zinnwald/Cinnovec mine (Germany). Combining hyperspectral and geometric data provides unique 3D hypercloud results interpreted for mineral
and structural features. Hyperspectral analysis successfully identifies iron and clay minerals along with multiple vein and fault structures. A remote mineral exploration method in underground mines significantly improves safety by keeping the operator away from hazardous areas. The proposed platform and workflow show potential to contribute to underground mineral exploration, especially if future improvements in data quality and autonomous capabilities are made.

Lithium is a highly sought-after resource currently. Among other things, its use in lithium-ion batteries used extensively in electric vehicles, mobile phones and other electronics keeps the demand high. The demand for lithium is even projected to double in the coming years. It is safe to say that lithium will continue to remain an important resource for the future. Because of this increasing demand, the supply of lithium has grow. This supply can increase by expanding current mines, or opening new ones. Lithium can be found in lithium-caesium-tantalum pegmatites, a rare subcategory of granitic pegmatites. In these pegmatites, the lithium-containing minerals spodumene, petalite and lepidolite are most commonly mined. These minerals can be found in the form of larger crystals, but in recent years more attention has come to lithologies where quartz and lithium-containing minerals have been found growing side-by-side in smaller crystals.
This thesis aims to propose two models to quantify the amount of lithium-bearing minerals in fine matrix. These models are created by using Fourier-Transform Infrared Spectroscopy (FTIR) data of powder mixtures consisting of petalite and quartz for the first model, and spodumene and quartz for the second model. This approach using powders allows for simple and fast data acquisition which aids model development. The models considered in this thesis are linear regression, ridge regression and support vector regression. Linear regression produces the best results. Both models have shown to accurately predict the respective weight percentages of petalite and spodumene in powder mixtures with quartz. The spodumene model, when applied to a rock which contains both spodumene and quartz in fine matrix, produces similar results compared to Laser-Induced Breakdown Spectroscopy (LIBS) and thin section analysis, which shows that FTIR models created using powder mixtures can be utilized for rock samples as well. To expand on the results presented in this thesis, further testing of the models presented here, and looking into models that can predict mixtures of three or more mixtures is recommended.